DE2319650A1 - METHOD AND ARRANGEMENT FOR CONVERTING A MESSAGE SIGNAL FROM ONE DIGITAL CODE FORMAT TO ANOTHER - Google Patents

Description

Procedure and arrangement; for converting a message signal from a digital code format to another

The invention relates to a method and an arrangement for converting a message signal from a digital code format to another.

Differential pulse code modulation (DPCM) is a form of message coding in which an analog speech or Video signal is sampled periodically to form a digital pulse train, and in which the difference between of each pulsed sample and a prediction of it based on the most recent samples for transmission is quantized and encoded. By using a number of quantized levels, the analog input signal becomes approximated in a stepped shape. Differential Coding by which the redundancy of the message signal is removed can be compared to conventional PCM coding lead to a bit saving of around two bits per sample.

309843/1096

However, if the quantizer is provided with steps of a fixed size is, it is not from the coded differential signal always efficiently occupied. Ideally, each step should be occupied with the same probability. Furthermore, a Differentiating encoders only follow rapid changes in the input signal to a limited extent. This limitation and the The associated coding error is too large Steepness returned. An encoder that works in this way is said to be position limited.

Too great a steepness is avoided to some extent by the fact that the quantifier reacts to the changing signal parameters is adjusted. An adaptive differential pulse code modulus. lation arrangement (ADPCM) in its simplest form monitors the digital output signal of the encoder and changes it by responding to pulse trains indicative of the magnitude of the difference signals, the effective step size of the quantizer. If the slope is too great, for example, then the encoder output signal is a sequence of pulse groups,. which indicate that a maximum increment is required. If the signal is very small, the output side jumps Pulse train typically between the levels of the lowest quantizer stages. In both cases speaks a logic control unit that monitors the output signal and sets the effective step size of the quanti-

309843/1096

sierers a. This can be done directly in the quantizer or by changing the level of the quantizer and decoder. In this way, the ADPCM code combats the problem of having too great a steepness, while at the same time retaining the advantages of DPCM coding. As a result, the transmitted code is more efficient because it allows more signal redundancy to be removed from the transmitted data. For the same bit rate, it delivers a signal of higher quality or, conversely, at a lower bit rate, it produces a signal of predetermined quality.

So there are several forms of message coding available, each different from the other Has advantages and disadvantages. Some messages get better in one and some messages get better in the other Coded shape. In the case of extensive transmission systems, however, it is imperative to have digital arrangements for to insert different code formats. For example, it may be desirable to have a conventional Transmit a PCM signal over an ADPCM channel to provide a ' to obtain advantageous bit rate. Or it may be that only one DPCM channel is available. Maybe adjacent channels for several different code formats are encountered during transmission. Because a channel that

309843/1096

is intended for one code format, cannot adapt a message encoded in another format, it is necessary to convert the message from one format to the other at each intermediate point.

It may also be desirable to convert one bit rate to another, for example a 3-bit ADPCM format to a 4-bit DPCM format, a 6-bit PCM format to a 3-bit ADPCM format, and so on . Furthermore, it is often necessary within half convert a code format to another a single array, so that a message signal, for example, can be digitally filtered ^first Most digital filters work in the PCM format. It is therefore necessary to filter what is to be filtered. To convert the message signal to PCM and then to convert it back to the original transmission format.

Previously, the conversion of message signals from one code format to another was local decoding, i.e., reducing into the analog form of a basic tape, and then recoding into the new format is necessary. E.g. becomes a ADPCM signal into a pulse amplitude modulated (PAM) signal decoded, demodulated in a low pass filter, then re-quantized and encoded as a DPCM message signal. Obviously, this is a complicated coding and decoding arrangement necessary. In addition, it favors coding

309843/1096

errors in the signal and often loads it with quantization noises and other signal distortions.

The aforementioned problem of converting a coded signal from one digital format to another is achieved according to the invention without signal distortion by converting on a purely digital basis without decoding into the baseband by logically checking the message signal that is in digital forwarding of a first format is encoded, takes place, to determine the conditions involved in removing the redundancy of the message signal during its Coding has been applied by digitally multiplying the digital words in the first format by a scale factor be chosen so that digital product. Words are formed in which the redundancy of the message signal is redistributed, and in which the digital product words to represent the message signal the number of bits selected for the second coding word must be rounded off.

The drawings show:

1 shows a conventional arrangement for digitally coding an analog signal according to known ones predictive quantization methods;

309843/1096

2 shows a typical step characteristic of a cuautizer and the 3-bit coding used to identify the different levels of the quantizer;

Fig. 3 is a block diagram of an arrangement that converts a digital ADPCM signal into both a digital DPCM and a also converts to a digital PCM signal;

Figure 4 is a block diagram of an arrangement that converts a digital DPCM signal into a digital ADPCM signal converts; and

Fig. 5 is a block diagram of an arrangement that generates a signal in a digital PCM format into a Signal.in converts to a digital DPCM format.

According to the invention, a coded message signal is converted from one digital code format into another on a purely digital basis without decoding into the baseband. In a few words, the essential difference between the various predictive code forms is how extensively and in what way the signal redundancy of a message signal is removed. Therefore, an "a priori" -Knowledge of the nature and extent is the redundancy removal used to restore the redundant information so far and / or redistribute it, as required to change the code of Forrnates Nächrichtensignales is.

309843/1096 "

For example, the logical coding unit of an adaptive coding arrangement is used to define the required setting range in order to scale a message signal and a quantizer. Usually the logic unit checks the bit stream at the encoder output and indicates that the input signal is at the highest or lowest quantization level. It responds by scaling the signal to match the quantizer, or by it changes the step sizes of the quantizer itself to classify the signal.

The conversion according to the invention is carried out in the same way s arrangement uses a logical unit that uses an incoming bit stream to identify a signal sequence checked and indicates that there is a scale factor in the signal coding was formed. The logical unit then sets a, taking this scale factor into account another scaling factor that is suitable for converting the signal into a new format. The format is through digital multiplication of each code word with the required scale factor and by rounding off the product word converted to the number of bits required for the new format. Discarded least significant bits are accumulated and translated into the new word to make the conversion more accurate.

309843/1096

In addition to converting from one digital format to the other, it can be used in a predetermined code format Word length changed, with or without changing the code format. For example, a 4-bit DPCM signal as a 3-bit signal in an ADPCM format or as a a 6-bit signal can be redetermined in a PCM format.

Because the format conversion to x ^ is carried out effectively to be able to, on an · "a prior!" - knowledge of the nature and Based on the manner in which the redundancy of a message signal was changed during encoding, it therefore serves as well as understanding the invention to discuss how digital predictive coding is carried out.

Differential PCM is a special form of predictive It is used to quantize the difference between a sample of an analog signal and a linear prediction to quantize from him. The difference signal, sometimes called the prediction error, requires for the transmission typically fewer bits than the quantized value of the original input sample. It was also shown that the noise in the determined Signal due to the prediction in the quantizer is the same as the quantization noise in the error signal. That

309843/1096

Quantization noise therefore does not accumulate with successive estimates of the input signal on the transmission side.

Fig. 1 shows an arrangement for coding analog message signals in either a DPCM or an ADPCM format. A message signal is in the unit 11 with a selected speed scanned by a clock or the like. is controlled and the samples are in the subtraction network 12 with the amplitude of Sample prediction compared. The difference or the prediction error signal arrives at the quantizer 13 typically fixed quantization levels to take signal samples at selected quantization levels. The quantized signals are sent to the linear prediction unit 15, which is generally an integrator, and to it serves, the output signal of the quantizer is stepped to hold, lqgt. The integrator delivers one from the last Sample values accumulated signal and therefore a reasonably accurate estimate of the value of an incoming Signal to the subtraction network 12.

The quantized signals are also sent to the encoder 16 in which a suitable pulse train code is set up to identify the pulse train values. The encoder

309843/1096

16 kanri, for example, digital 3 ~ bit words in a conventional one Deliver PCM format. The coded signals are then digitally transmitted via channel 17 to a receiver arrangement transmitted, where incoming signals are decoded in the unit and passed on to the linear prediction unit 19 will. The prediction unit 19 corresponds in all respects to the transmission-side prediction unit 15 and is as these are typically implemented as an integration network. Incoming samples are, in the prediction unit 19 accumulated, smoothed in the filter 20 and as a mapping of the input-side message signal to a circuit on the output side. Because the arrangement has a differentiating effect, fewer bits are required for specification of the applied message signal than required without predictive quantization.

An adaptive DPCM arrangement works essentially with it the same elements, but adjusts the effective quantization characteristic to the local Detect difference signal regardless of large changes in scale. In ADPCM mode, the output-side Bit stream of the encoder 16 is monitored in the logic unit 22 in order to determine those code words which show the highest or lowest level occupancy of the quantizer. When the highest level for a selected

3 09843/1096

Interval are occupied with too great a slope, causes the logical unit is expanding the quantizer, i.e. changing the approximate step sizes by the Adapt signal with high amplitude. When code words from encoder 16 indicate that the lowest level of the Quantizer is occupied for a specified interval with a low level, causes the logical unit a compression of the quantizer to reduce granular distortion. Such a one Adaptation by compression or expansion, commonly referred to as companding, can be either at the speed of syllables (long-term) or at the moment (short-term). The companding is usually obtained by a multiplicative change in the reference levels of the quantizer 13 and at the transmission end of the receiver-side decoder 18 is reached. Consequently are adaptive arrangements by selectively changing the step size to changes in the amplitude of the applied Signal marked out.

In an ADPCM arrangement, signals sent to the receiver run, both to the decoder 18 at the receiving end and to the logic unit 23 at the receiving end. The logic unit 23 is essentially with the transmitter side Logical unit 22 is identical and has access like this

309843/1096

to the same bit stream and makes the same decisions as this. The logical Unit 23 receives a signal (a multiplication signal) representing the reference voltage range of the decoder 18 upon conversion sets incoming digital signals into an analog signal. So if incoming digital signals display the occupancy of the highest quantizer level, , the logic unit 22 sets the range of the quantizer 13 entirely accordingly. The logical unit 23 works also in this way on the decoder 18, so that a greater voltage than the analog counter-image of this digital Signal is set. If the incoming digital signal shows the occupancy of the lowest quantizer level, then the logic unit scale factor applied to decoder 18 will similarly reduce the size of the analog signal assigned to this digital word.

In general, the logical unit scale factor can be any number of rules between instantaneous and syllable adjustment with a large memory

■ follow. Fig. 2 shows a typical quantization staircase with their 3-bit step code contained in an ADPCM speech coding arrangement is used. If the binary code word is 111 or 000, a signal with a large amplitude indicates then a multiplication factor that

309843 / 1-096-

is greater than 1 is used to expand the effective signal size of the quantizer. If the code word is 100 or 011, a multiplication factor slightly smaller than 1 is used to reduce the signal size of the quantizer second hand. A logical algorithm suitable for two different sampling frequencies is shown below in table. 1 listed.

Table I. Word at the coding output Multiplier unit at the output

the logical unit

8KHz sampling 6 KHz sampling

111 or 000 1.750 - 1.625

110 or 001 1.250 1.250

010,011,100,101 0.875 0.800

Fig. 3 shows an arrangement for converting an ADPCM bit stream, which is prepared as in the arrangement shown in FIG. 1, in a DPCM bit stream. Such Working method may be required in a transmission arrangement be, for example, for digital filtering, for operating a computer or the like an interface takes place with another arrangement or within an arrangement. As an example, consider an arrangement with a 3-bit ADPCM word on the input side explained. Incoming

309843/1096

Ϋ / words are passed to the logical unit 30 which they checked "and a suitable binary multiplication factor forms to expand or compress the scale factor of the applied signal. The logical unit 30 is with that used to 'prepare the ADPCM bit stream logical unit 22 is identical and forms multiplication factors, which match the multiplication factors formed by the logical unit. You v / ground used here to scale the incoming signal. The digital multiplier 31 of any design responds to two η-bit signals (e.g. η = 3) and generates a 2n-bit signal on the output side (6-bit in this case Example). Digital multipliers are well known and widely used. By removing the when quantizing The expansion and compression factor used for the ADPCM message signal is the output signal of the multiplier 31 a DPCM encoded signal. In the present example it is expressed in 6-bit words. The 6-bit Product signal is stored in product register 32.

However, it may be that interconnection circuits operating in the DPCM format "(corrarting circuits) no 6-bit words can process. It is better if the number of bits is only 1 or 2 larger than in the ADPCM format, e.g. 4-bit

309843/1096

amounts to. The product register 32 is set up to .Product signals in an arrangement according to their bit value save. A 4-bit signal is therefore generated by rounding off the register, i.e. by reading out only the four most significant Bits of the 6-bit product generated. When the output signal of register 32 is to be a 5-bit signal, the 5 most significant bits (MSB) are read out.

Although the rounding is done to get a sharp image of the To form the desired signal in the new format, it is obvious that rounding-off errors can arise. Consequently the Least Significant Bits (LSB), instead of being discarded, become accumulator 33 and after being temporarily stored led back into the output bit stream.

The accumulator 33 is used to record every digital code value, e.g. in the delay unit 34 (delay 34) and digitally adds it to the next following digital code word in adder 35 · every time the accumulated Sum is 1, a digital code value is read out, leaving some fraction for further accumulation. Digitally working accumulators are known and are usually used for digital caching.

309843/1096

The rounded output of register 32 is on this way along with any buffering signals from the accumulator 33 to the adder 36. The summed up digital output signal is the desired one DPCM signal.

When the desired output signal is a conventional PCM signal is, the DPCM signal from the adder 36 becomes the accumulator 37 led. The accumulator 37 serves as a digital decoder with the receiver-side shown in FIG Decoder 18 is comparable. Its digital output signal is a conventional PCM signal, and As with the DPCM signal, its word length is determined by the scope the rounding of the scaled input signal and determined by the adding accuracy of the accumulator.

In addition to the. Conversions from ADPCM to DPCM are often conversions as well required in the opposite direction. The latter are a little more difficult to implement because of the redundancy involved in coding removed rather than built in, i.e. adaptive coding is necessary. According to the invention, however additionally encoded on a fully digital basis and the bit stream size selectively adjusted to the desired one Adapt ADPCM format.

309843/1096

4 shows an arrangement for the digital conversion of DCPM to ADPCM. An incoming digital DPCM signal in a 4-bit code format, for example, is converted into PCM with the aid of an accumulator 41. Corresponding the sequence already described previously in connection with the accumulator 37, which is shown in FIG. 3. Of the Accumulator 41 has an addition register, and therefore its output signal can be adjusted to any desired accuracy rounded down, for example to an accuracy of 6 bits. The PCM signal on the output side is digital Subtracter 42 is compared to a local digital word estimate and a digital difference signal generated. The local estimate must necessarily be of the type and amount of companding desired for the ADPCM format reproduce. Because the coding is digital and the multiplication is digital the bit stream size changes, according to the invention registers are provided for the accumulated signals, and the signals are rounded down to produce the desired bit stream size. The example uses a 6-bit signal applied from accumulator 41 to subtracter 42, a local 6-bit signal estimate from the accumulated Signal extracted and a 6-bit difference signal applied to register 43. If the output side For example, the ADPCM signal is expressed as a 3-bit signal

309843/1096

is to be, the register 43 rounds off the stored signal and supplies the adder 44 with a sequence of 3-bit signals. In order to make the output signal more precise, the least significant bits are temporarily stored in the accumulator 45 made, in this case the discarded three least significant Bits until a bit weight is accumulated and then the signal stored in adder 44 is incremented. The output of adder 44 represents the ADPCM signal in the form of a 3-bit digital word.

The local estimate of the incoming PCM signal becomes as in predictive coding by integrating a processed version of the output signal. The adaptive signal characteristics required for ADPCM. the output signal is obtained by changing its scale factor is set in multiplier 47, for example. The multiplication factor is determined by the logical Unit 46 is formed, the output bit stream, for example with the help of a previously described and the algorithm shown in Table I on quantizer occupancy characteristics checked.

The scaled signal is then applied to the product register 48. When the logical unit 46 is a

30384371096

3-bit multiplication signal supplies and the output-side bit stream of the adder 44 is a 3-bit signal, then forms the multiplier 47 is a 6-bit signal. The registry

48 must therefore be able to accommodate at least 6 bits. When the local signal estimate is formed with an accuracy which is less than 6 bits, the digital words in register 48 can be rounded off in the form described above will. The signal set to scale and stored in register 48 then arrives at the accumulator 49, the acts like a prediction unit and is designed for 6-bit accuracy. The output of the accumulator

49 represents a local estimate of the PCM signal on the input side and is used in the subtracter 42 from the incoming PCM bit stream subtracted.

Because the DPCM signal is formed by a prediction unit that removes the portion from each sample of a PCM signal that is based on a previous signal representation can be predicted, it is only necessary to scan an incoming PCM sample by a predictable increment to convert it to a DPCM sample. The PCM samples in Fig. 5 therefore become Difference network 50 led, which can be designed arbitrarily ' can. Typically it has a subtracter 51 and a sample delay unit 52. Every other

309843/1096

The next sample is reduced in the subtracter 51 by the sample of the immediately preceding sample. This difference forms a DPCM signal, and DPCM words of the same length are formed if the subtracter has an accuracy equal to that in the input-side PCM signal
As an alternative to this, the length of the DPCM word can be reduced by rounding off in the subtracter 51, which can also be carried out better.

309843/1096

Claims (5)

Claims Method for converting a message signal from one digital code format to another, characterized in that the conversion is on a purely digital basis without decoding into baseband (in 30) by logically checking the message signal, which in digital words is a The first format (ADPCM) is coded to determine the conditions involved in the removal of the Redundancy of the message signal during its coding have been applied that the digital words in the first format with a scale factor digital (in 31) chosen so that digital product words are formed in which the redundancy of the message signal is redistributed, and that the digital product words for representing the message signal are rounded down to the number of bits selected for the second coding word (in 32). 2. Facility for carrying out the procedure according to ■ Claim 1, with a digital system for converting a digital message signal represented in ADPCH into a digital signal in DPCM, 309843/1096 characterized by a logic network (30) based on the companding characteristic of an applied ADPCM signal to produce a digital scale factor representative of the effective ones used in predictive coding of a signal Represents quantization characteristic, a multiplier network (31) for digital multiplying of each ADPCM signal word with the scale factor to form digital product word, a product register (32) for selecting a predetermined number of the most significant Bits of each digital product word, an accumulator network (33) to retrieve the least significant Bits, successive product words until a bit weight is accumulated, and an adder network (36) for adding accumulated bits to the selected digital product words to obtain a corrected Form the product word as a DPCM representation of the applied signal. 3. Device according to claim 2, characterized, that the DPCM representation of the applied signal is accumulated word for word (in 37) to form a PCM representation of the applied signal. 309843/1096 4. Device for performing the method according to claim 1 with a digital circuit arrangement for converting a digital message signal represented in DPCM into a digital signal in ADPCM, characterized by an accumulator network (41) for digitally accumulating successive words in the DPCM code by the earlier ones to restore redundancy removed during coding, an accumulator (49) for generating a digital one Signal that reconstructs the code words based on an associated companding characteristic predicts, and a network (43) to register a digital difference between the restored and predicted signal and the selective one Rounding off the difference signal in order to obtain a digital ADPCM signal with a given digital format form. 5. Method for converting a message signal from ^% one PCM format to another, characterized by reducing an incoming, represented in PCM code Signal by a predictable increment in order to convert the signal into a signal represented in the DPCM code. 309843/1096